VASC Seminar Announcement
            =========================

Date: Monday, 2/21/00
Time: 3:45:4:40
Place: NSH 3002

Speaker:  Pierre Lamon
Title: Fingerprint sequences for mobil robot localization

The three most important questions in mobile robotics are: Where am I?
Where am I going? and How do I get there? In this work, we address the
first question of mobile robot localization. Most of recent works
solve this problem by either map matching or landmark detection. These
approaches work well in indoor environments, while suffer drawbacks in
the outdoor environments. As a solution to these drawbacks,
vision-based localization method has recently witnessed a newfound
popularity.  The CCD Camera is a popular choice for mobile robot
sensing because it is not inherently dependent on environmental
geometry like ranging devices.  Therefore, it is hoped that a
transition to indoor and outdoor navigation will be more
straightforward with vision.

As the fingerprints of a person are unique, so are the visual
characteristics at each location.  The thesis of this localization
system is that: 1) a unique virtual fingerprint of the current
location can be created, and 2) the actual location of a mobile robot
may be recovered by comparing this fingerprint to a database of known
and identified fingerprints. We define a fingerprint as a circular
list of features, which consists of a sequence of vertical edges and
color patches.

To achieve localization we use the following strategy: first we
compute fingerprints sequences for a sample of evenly spaced locations
in the environment. We call them map points. During the navigation,
the robot computes the fingerprint of its current position and finds
the most similar map points.  From this match, we can deduce the
location of the mobile robot.